| title: “Geometric Morphometric Analysis of Projectile Points from the Southwest United States” |
| author: |
| - firstname: Robert J. |
| lastname: Bischoff |
| affiliation: 1 |
| address: |
| - code: 1 |
| address: School of Human Evolution and Social Change, Arizona State University, Tempe, AZ |
| email: rbischoff/@asu.edu |
| abstract: | |
| Traditional analyses of projectile points often use visual identification, the presence or absence of discrete characteristics, or linear measurements and angles to classify points into distinct types. Geometric morphometrics provides additional tools for analyzing, visualizing, and comparing projectile point morphology utilizing the whole or parts of the form in either two or three dimensions. This study is an analysis of the effectiveness of geometric morphometric methods for identifying technological similarity in 2D projectile point outlines for previously classified late prehistoric projectile points found in the U.S. Southwest and unclassified projectile points from Tonto Basin, Arizona. Various methods from geometric morphometrics were compared to determine which method best reproduced the original classification scheme. Elliptical Fourier analysis was compared with various configurations of semilandmark and landmark analyses using generalized Procrustes analysis. These methods were applied to the complete projectile point form, and the landmark analysis was also applied to half of the lower quadrant of the projectile point—essentially one corner of the projectile point. The landmark analysis applied to the corner of the projectile point provided the best results. This method was then applied to the Tonto Basin points. Hierarchical clustering was used on the Tonto Basin projectile point morphometric data to explore the variation in shapes between sites. To demonstrate that geometric morphometric methods can be used without relying on typologies, a network analysis of the morphometric distances was also conducted. This network graph produced distinct clusters of technological similarity in projectile point outlines, while also showing the continuous variation between points. These results demonstrate the effectiveness of geometric morphometrics for the 2D analysis of late prehistoric arrow points in the U.S. Southwest. |
| keywords: |
| - American Southwest |
| - Hohokam |
| - Arizona |
| - projectile points |
| - lithics |
| - computational archaeology |
| - geometric morphometrics |
| caption-side: bottom |
| bibliography: references.bib |
| biblio-style: journal-of-computer-applications-in-archaeology.csl |
| link-citations: yes |
| editor_options: |
| chunk_output_type: console |
| output: |
| bookdown::pdf_document2: |
| keep_tex: true |
| fig_caption: true |
| template: template/main.tex |
| citation_package: biblatex |
| urlcolor: blue |
Geometric morphometrics (GM) is a quantitative approach to studying shape in two or three dimensions that has recently been adopted in archaeology [see @MacLeod2017-yl; @Okumura2019-ur; @Shott2010-fn for overviews]. It has numerous advantages over traditional lithic analyses, particularly because it can overcome the reliance on linear dimensions [@Shott2010-fn, 196-197]. Lithic artifacts can be assigned to typologies or directly compared without the use of a typology, as will be demonstrated in this paper. There are several approaches within GM that provide similar results through different methods. One of the more traditional approaches is to place landmarks at homologous locations around the object. Landmarks can be augmented with semilandmarks, which are points placed relative to another using a consistent rule—usually equidistant spacing between two points [@Okumura2019-ur, 2-4]. Another common approach is to use elliptical Fourier analysis to compare the outlines of objects. Each method has strengths and weaknesses. A major purpose of this study is to evaluate the effectiveness of these methods for analyzing projectile points in the U.S. Southwest during the late prehistoric period (specifically during the Hohokam Classic Period—AD 1100-1500).
Once the method of analyzing the projectile points has been determined, the next step is to determine how to compare projectile points using the results of the analysis. One approach would be to use an existing regional typology and to assign projectile points to the closest match [e.g., @Kocer2017-au]. Another approach, would be to use cluster analysis to assign projectile points to newly created types [e.g., Petrik2018-pd; Matzig2021-id]. The final approach would be to ignore typologies and compare the morphometric distance for each projectile point directly. This is the second primary purpose of this study—to evaluate the effectiveness of these approaches for use in analyzing projectile points from the Southwest United States.
Regional analyses are fundamental parts of archaeology, but there are many challenges to overcome. One of these challenges is harmonizing the different categorization schemes (i.e., ontologies) used throughout the region. Another of these challenges, is determining whether the current categories are useful. The U.S. Southwest has a long history of regional ceramic typologies [e.g., @Colton1956-zy; @gladwin1930a; @Hargrave1932-ng; @Kidder1915-ae; @Martin1940-jg], but there are still disagreements, challenges, and competing definitions [@Duff1996-au]. Regional analyses in the Southwest, based in large part on pottery, have produced many useful insights [e.g., @Bernardini2005-ue; @Clark2019-bz]; @Hegmon2016-xw; @Mills2013-wq; @Peeples2018-ib]. However, one type of material culture that has received little attention—in the Southwest at least—is lithics (i.e., chipped stone). Projectile points are commonly discussed during the archaic period of the Southwest, and they are common topics in many other areas of the North American continent and world where they are found, but they are rarely discussed after the appearance of pottery.
Despite the over-emphasis on pottery in the Southwest, there are some excellent resources on projectile point typologies [e.g., @Hoffman1997-hb; @Justice2002-cf; @Loendorf2004-tp]; @Sliva2006-nq]. However, ad hoc approaches are common, and these cannot easily be extrapolated beyond specific projects. Even using existing resources can make comparisons difficult. How does Tagg’s [-@Tagg1994-wi, p.111] Type 23 compare to Sliva’s [@Sliva2006-nq, p. 35] Cohonina Side-notched? There is an answer, but often it is easier to come up with a new typology schema than to try to harmonize existing work.
Another challenge that is not unique to projectile points is that interpretations may differ between analysts. Exactly when does a base begin curving enough to be called basal notched? Even the difference between a side-notched and a corner-notched point can, at times, be ambiguous. Not to mention the frustrating situation where a point appears to have one corner-notch and one side-notch. How should one place this point into an existing typology? These are questions that can be handled in different ways that differ from analyst to analyst. Idiosyncrasies and biases are impossible to be rid of entirely, but using approaches such as those described in this paper can reduce them and increase the reproducibility of the process.
By necessity, this paper covers a number of topics. The geographic area is the U.S. Southwest, but the methods and analysis are applicable to any area. The primary purpose was to explore geometric morphometric methods using previously typed specimens from the Southwest and untyped specimens from the Tonto Basin. Another purpose was to analyze the results with and without using typologies. The results demonstrate that, in this particular case, a combined landmark and semilandmark approach is most effective and that useful analyses can be conducted with and without the use of typologies.
In order to test the effectiveness of geometric morphometric methods, I needed a dataset of well-typed projectile points that could be used as a validation set. I chose to use the typology published by Noel Justice [-@Justice2002-cf] for the simple reason that it is easily accessible and contains numerous illustrations. These illustrations were used as type specimens to compare projectile points from Tonto Basin in central Arizona (Figure 1). These points were excavated in a series of large cultural resource management projects necessitated by work on the Roosevelt Dam. The largest project—the Roosevelt Platform Mound Study—included 129 sites. Most of the sites date between AD 1275 and 1325 with occupation continuing until around AD 1450 [@Rice1998-ku]. In the original analysis, projectile points were classified according to small and large points and then subdivided based on morphological characteristics [@Rice1994-rk, p.727]. The typology used is an excellent demonstration of the difficulty in conducting projectile point studies in this area, as the typology is idiosyncratic to this specific project, and cannot be easily compared with other datasets. This is not a criticism of the analyst’s choice to create a new typology, as no existing typology met the needs of the researchers.